Accepted_test

This study aimed to try apply methods of artificial neural network for prediction in silico potential ligands for CD8+ T-cell epitopes of MHC class I in pig model using freely available machine leaning algorithms. Salmonella enterica subsp. enterica serovar Choleraesuis, the causative agent of salmonellosis, that is known as a swine-adapted foodborne pathogen associated to invasive infections in both animals and humans was used as a model pathogen. For this purpose, three different neural networks, such as conventional sparse encoding, BLOSUM encoding and hidden Markov model encoding, the main tools for T Cell Prediction - Class I were applied for prediction of immunogenic epitopes for proteins encoded by 14 Salmonella virulence genes (ipfB, sopB, ompA, pipB, pipB2, sifA, sinH, sopD, sopD2, sopE2, sseB, spvB, spvC and spvR) addressed to 44 available pig alleles. Overall, in range 60,000 - 400,000 epitopes were predicted for each protein. Two consecutive cut-offs resulted for individual protein from 14 to 88 epitopes from which only 20 potentially protective epitopes associated with 5 proteins encoded by such genes as pipB, sinH, sopD, spvB and spvC were found. The data obtained could be perspective for development of multi-epitope peptide vaccine against porcine salmonellosis.